Pith. sign in

REVIEW 4 major objections 7 minor 48 references

Mitigating Group-Level Fairness Disparities in Federated Visual Language Models

T0 review · 4 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read FVL-FP reduces demographic disparity in federated vision-language models by an average of 45% while keeping task accuracy within 6% of state-of-the-art results.

desk verdict Plausible method and honest ablation work, but the headline fairness metric is undefined and internally inconsistent, so the central claim is not yet supported. read the letter →

arxiv 2505.01851 v1 pith:CNMCHIE6 submitted 2025-05-03 cs.CV

Chaomeng Chen , Zitong Yu , Junhao Dong , Sen Su , Linlin Shen , Shutao Xia , Xiaochun Cao This is my paper · ORCID
classification cs.CV
keywords federatedlearninggroupfairnessvision-languagemodelsprompttuningorthogonalprojectiondemographicbiasnon-IIDdata
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes FVL-FP, a prompt-tuning framework that tackles group-level fairness in federated vision-language models without centralizing user data. It claims to cut demographic disparity by an average of 45% compared with standard federated learning, while keeping task accuracy within 6% of state-of-the-art results. The method works by training small fairness-aware prompts locally, projecting demographic bias out of image representations, and fusing client prompts with weights that reward both accuracy and fairness. A sympathetic reader should see this as evidence that fairness in privacy-preserving multimodal systems can be improved by a parameter-efficient, interpretable mechanism rather than costly retraining.

What carries the argument

The load-bearing mechanism is the Demographic Subspace Orthogonal Projection (DSOP). It encodes demographic prompts such as “a photo of a man” and “a photo of a woman” through the frozen text encoder, collects the embeddings into a matrix, takes the top-k singular vectors as the demographic subspace $V_a$, and computes the debiased image representation as $z_{\text{debiased}} = z - \mathrm{Proj}_{V_a}(z)$. This carries the argument because it gives an interpretable geometric way to strip protected-attribute information while retaining task-relevant content; a fairness-aware contrastive loss then penalizes any residual cosine similarity to demographic prompts. The other two components support this core: CDFP supplies fair prompts locally, and FPF reweights client contributions during aggregation.

What would settle it

Train or evaluate the framework on data where bias is generated by a nonlinear function of the protected attribute (for example, gender affects only high-order feature interactions), or swap the demographic attribute in paired images and check whether the projected representations of the two versions remain identical. If DSOP's fairness gains disappear or accuracy collapses under nonlinear bias, the linear-subspace assumption is the failure point.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that demographic bias in federated vision-language models can be substantially reduced by prompt tuning alone, without updating the underlying vision-language model. The framework combines three mechanisms: CDFP inserts demographic fair prompts across transformer layers with adaptive residual connections to neutralize biased attention; DSOP constructs a low-dimensional demographic subspace from text embeddings of demographic prompts and orthogonally projects image representations away from it; FPF aggregates client prompts by weighting each client by accuracy times one minus measured bias. Together these yield the reported 45% average reduction in demographic disparity across CelebA and FairFace smiling and age detection tasks, with balanced accuracy within about 6% of state-of-the-art federated baselines.

Load-bearing premise

The method assumes that demographic bias in the model's image representations lies in a low-dimensional linear subspace spanned by text embeddings of demographic prompts, so subtracting that subspace removes bias without removing task-relevant information.

Editorial extensions

If this is right

  • Only prompt vectors are updated and exchanged, so the fairness gains avoid retraining the billion-parameter backbone and add minimal communication overhead.
  • Demographic disparity falls by an average of 45% versus standard federated approaches while balanced accuracy stays within about 6% of state-of-the-art baselines.
  • The method remains effective as data heterogeneity increases (Dirichlet $\alpha$ down to 0.1) and as the number of clients grows to 40, retaining roughly 96–98% of centralized-training accuracy.
  • Improvements are larger on harder tasks: on age detection, fairness disparity drops by up to 87% on CelebA and 83% on FairFace.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If demographic bias in the embeddings really is near-linear, DSOP could be used as an inference-time post-hoc debiaser on off-the-shelf models, which would extend the contribution beyond federated settings.
  • The subspace is built from a fixed set of demographic prompt texts, so results are likely sensitive to prompt wording and attribute coverage; testing many prompt templates would reveal how much of the gain depends on that choice.
  • The paper evaluates single protected attributes (gender) on face datasets; extending to intersecting attributes such as gender × race would stress whether one low-rank subspace can capture multiple entangled biases.
  • A direct test of the mechanism's specificity would be to replace the demographic subspace with a random subspace of the same rank; if fairness gains persist, the gains would come from regularization rather than identification of bias directions.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes FVL-FP, a federated prompt-tuning framework for group fairness in vision-language models. It combines three components: Cross-Layer Demographic Fair Prompting (CDFP), Demographic Subspace Orthogonal Projection (DSOP), and Fair-aware Prompt Fusion (FPF). The authors claim an average 45% reduction in demographic disparity relative to standard FL methods while maintaining task performance within 6% of state-of-the-art results, based on experiments on CelebA and FairFace for smiling and age detection tasks, together with ablation, data-heterogeneity, and client-count studies.

Significance. If the claims were substantiated, FVL-FP would be a useful parameter-efficient contribution to federated VLM fairness, since it avoids full model retraining and provides a geometric, interpretable debiasing mechanism. The paper includes mean/std statistics, comparisons to several FL fairness baselines, ablations of each module, and robustness studies over heterogeneity and number of clients. However, the headline quantitative claim is currently not verifiable because the main fairness metric used in the tables is undefined and the dataset count is misstated; the significance is therefore conditional on a corrected and reproducible evaluation.

major comments (4)
  1. [Section 3, Eq. (1) and Tables 1, 3, 5] The fairness metric Φ_A is never defined, and the values reported for age detection violate the only formal definition in the paper. Eq. (1) defines global fairness as an absolute difference of true positive rates, which is bounded by 1; Table 1 reports Φ_A = 1.829 for CLIP zero-shot and 1.898 for FedAvg on CelebA age detection, and Table 3 reports values up to 2.195. Because the abstract's 45% reduction is computed from these Φ_A values, the central claim lacks a verified referent. The authors must define Φ_A (and Φ_demo, Φ_eq) and either correct the values or replace them with a metric consistent with Eq. (1).
  2. [Section 5.1, Implementation Details, Tables 1 and 5] The experimental conditions for the headline results are inconsistent. Section 5.1 says all methods are compared under 5 clients, while Implementation Details says the system consists of four nodes and the default number of nodes is 4; moreover, Table 1's FVL-FP entries exactly match Table 5's N=20 row (e.g., A_B=0.915, Φ_A=0.139, Φ_demo=0.008, Φ_eq=0.023 for smiling detection). Please state precisely which client count produced Table 1 and reconcile the text.
  3. [Section 5.1 and Appendix D] The implementation details needed to reproduce the method are missing from the manuscript. The referenced appendices are not present, and the main text does not give values for K (number of fair prompt basis vectors), k (top singular directions), the margin μ in Eq. (11), the trade-off weights λ and λ1, or the client data distributions. Since the method's behavior depends on these choices, the experiments are not reproducible as written.
  4. [Section 4.4, Eqs. (15)-(16) and Section 5.2] It is unclear whether the validation set used for FPF weighting and hyperparameter selection is disjoint from the balanced test set used to report Table 1. If the same data are used for both, then the reported fairness improvements are in part optimized on the evaluation set. Please specify the split and, if they are not disjoint, re-evaluate on a held-out test set.
minor comments (7)
  1. [Section 5.1] The phrase "As mentioned in," before the claim about smiling detection lacks a citation or reference.
  2. [Abstract and Conclusion] The abstract and conclusion claim evaluation on "four benchmark datasets," but Section 5.1 states that only CelebA and FairFace are used, and no fourth dataset appears in any table.
  3. [Section 4 heading] The section heading "Methology" should be "Methodology."
  4. [Table 2 and Section 4.4] Table 2 uses the acronym "FAPF" while Section 4.4 and elsewhere use "FPF"; please make the acronyms consistent.
  5. [Tables 1, 3, 4, 5] The table captions should define Φ_A, Φ_demo, and Φ_eq; currently these metrics are never defined in the text.
  6. [Table 1] The "FVL-FP (centralized)" column in Table 1 is not described in the experimental setup; please state how the centralized version is trained and how it differs from the federated variant.
  7. [Tables 1 and 5] Table 1 reports A_B=0.915 for FVL-FP on smiling detection, while Table 5 reports A_B=0.924 for N=5 and A_B=0.915 for N=20; the relationship between these two tables should be explained explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found; the reported fairness improvement is an empirical optimization outcome evaluated on a held-out test set, and the undefined Φ_A metric is a correctness/verifiability issue, not a circularity issue.

full rationale

FVL-FP is an empirical method. The paper optimizes fairness losses (Eqs. 11, 13, 18) and uses validation-based aggregation weights (Eqs. 15-16), then reports fairness and accuracy on a separately sampled balanced test set of 5000 images (Sec. 5.1). Reporting improvements on the quantity being optimized is standard supervised evaluation rather than a derivation-by-construction; no fitted parameter is renamed as an independent prediction, and no external benchmark is internally generated. The reference list contains no load-bearing self-citations: DSOP cites the independent INLP work (Ref. [25]) and compares against external baselines including FedAvg, FedProx, FairFed, and FF-DVP. The paper does, however, have serious verifiability problems that are outside circularity: Φ_A is never defined, and values such as 1.829 in Table 1 are impossible under Eq. (1)'s EOD definition (a difference of true positive rates is bounded by 1), so the headline "45% reduction" lacks a well-specified referent; the abstract says "four benchmark datasets" while Sec. 5.1 names only CelebA and FairFace; and the setup text alternates between five clients, four nodes, and Table 5's N=20 row matching Table 1's FVL-FP entries. These inconsistencies should be corrected, but they do not establish that the derivation is circular.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

No new physical entities are introduced. The method relies on trainable prompt vectors and geometric constructs (demographic subspace) that are internal to the algorithm. The ledger lists the key tunable parameters and domain assumptions that the central claim depends on.

free parameters (5)
  • fairness trade-off λ (Eq. 2) = grid searched, not reported
    Balances task loss and fairness loss in the local objective; value chosen by grid search on validation.
  • fairness regularization strength λ₁ (Eq. 13) = grid searched, not reported
    Controls the weight of the demographic contrastive fairness loss; tuned on validation.
  • margin μ (Eq. 11) = grid searched, not reported
    Upper bound on allowed similarity between debiased representation and demographic prompts; chosen by grid search.
  • top-k singular directions (Sec. 4.3.1) = not reported
    Dimension of demographic subspace used for orthogonal projection; choice affects how much information is removed.
  • number of fair prompt basis vectors K (Sec. 4.2) = not reported
    Number of demographic prompt vectors inserted at each layer; tuned by the authors.
assumptions (3)
  • domain assumption Demographic bias in CLIP visual representations is contained in a low-dimensional linear subspace spanned by text embeddings of demographic prompts.
    DSOP constructs V_a via SVD of text prompt embeddings (Sec. 4.3.1) and projects visual features off this subspace; if bias is nonlinear or outside this span, projection cannot remove it.
  • domain assumption Orthogonally projecting visual features away from the demographic subspace preserves task-relevant semantic content.
    The task loss LVLM (Eq. 12) is assumed to remain effective after debiasing, but the paper does not prove that task information is linearly separable from demographic information.
  • domain assumption Fairness measured on a server-side validation set can be used to reweight client prompts and transfers to the test distribution.
    The FPF aggregation uses Score based on validation accuracy and bias (Eqs. 15-16); this assumes validation bias is a reliable proxy for global fairness on the test set.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mitigating Group-Level Fairness Disparities in Federated Visual Language Models." pith.science (2026). https://pith.science/paper/CNMCHIE6

@misc{pith2026250501851,
  author       = {Pith},
  title        = {Pith review of: Mitigating Group-Level Fairness Disparities in Federated Visual Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CNMCHIE6}},
  note         = {Machine review of arXiv:2505.01851}
}
read the original abstract

Visual language models (VLMs) have shown remarkable capabilities in multimodal tasks but face challenges in maintaining fairness across demographic groups, particularly when deployed in federated learning (FL) environments. This paper addresses the critical issue of group fairness in federated VLMs by introducing FVL-FP, a novel framework that combines FL with fair prompt tuning techniques. We focus on mitigating demographic biases while preserving model performance through three innovative components: (1) Cross-Layer Demographic Fair Prompting (CDFP), which adjusts potentially biased embeddings through counterfactual regularization; (2) Demographic Subspace Orthogonal Projection (DSOP), which removes demographic bias in image representations by mapping fair prompt text to group subspaces; and (3) Fair-aware Prompt Fusion (FPF), which dynamically balances client contributions based on both performance and fairness metrics. Extensive evaluations across four benchmark datasets demonstrate that our approach reduces demographic disparity by an average of 45\% compared to standard FL approaches, while maintaining task performance within 6\% of state-of-the-art results. FVL-FP effectively addresses the challenges of non-IID data distributions in federated settings and introduces minimal computational overhead while providing significant fairness benefits. Our work presents a parameter-efficient solution to the critical challenge of ensuring equitable performance across demographic groups in privacy-preserving multimodal systems.

Figures

Figures reproduced from arXiv: 2505.01851 by the authors.

Figure 1
Figure 1. Federated Vision-Language Models. For each node, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 1
Figure 1. When deployed in federated environments, these biases [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The framework of FVL-FP. For each node, clients optimize for fairness through CDFP and DSOP algorithms. Following [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (2 more)
Figure 3
Figure 3. Figure 3: Cross-Layer Demographic Fair Prompting Algo [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]
Figure 4
Figure 4. Figure 4: Demographic Subspace Orthogonal Projection. We [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 30 canonical work pages

  1. [1]

    Jean-Baptiste Alayrac and et al. 2022. Flamingo: a Visual Language Model for Few-Shot Learning. arXiv preprint arXiv:2204.14198 (2022)

    arXiv 2022

  2. [2]

    Soumya Barikeri, Anne Lauscher, Ivan Vulić, and Goran Glavaš. 2021. Reddit- Bias: A real-world resource for bias evaluation and debiasing of conversational language models. arXiv preprint arXiv:2106.03521 (2021)

    arXiv 2021

  3. [3]

    Californians for Consumer Privacy. 2020. California Consumer Privacy Act Home Page. https://www.caprivacy.org/. Online; accessed 09-May-2022

    work page 2020

  4. [4]

    Tianshi Che, Ji Liu, Yang Zhou, Jiaxiang Ren, Jiwen Zhou, Victor Sheng, Huaiyu Dai, and Dejing Dou. 2023. Federated Learning of Large Language Models with Parameter-Efficient Prompt Tuning and Adaptive Optimization. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 7871–7888

    work page 2023

  5. [5]

    Yi Chuang and et al. 2023. Debiasing Federated Learning: Challenges and Op- portunities. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    work page 2023

  6. [6]

    Daniel de Vassimon Manela, David Errington, Thomas Fisher, Boris van Breugel, and Pasquale Minervini. 2021. Stereotype and skew: Quantifying gender bias in pre-trained and fine-tuned language models. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. 2232–2242

    work page 2021

  7. [7]

    Chenhe Dong, Yuexiang Xie, Bolin Ding, Ying Shen, and Yaliang Li. 2023. Tunable Soft Prompts are Messengers in Federated Learning. In The 2023 Conference on Empirical Methods in Natural Language Processing

    work page 2023

  8. [8]

    Wei Du, Depeng Xu, Xintao Wu, and Hanghang Tong. 2021. Fairness-aware Agnostic Federated Learning. In Proceedings of the 2021 SIAM International Conference on Data Mining, SDM 2021, Virtual Event, April 29 - May 1, 2021, Carlotta Demeniconi and Ian Davidson (Eds.). SIAM, 181–189. https://doi.org/ 10.1137/1.9781611976700.21

    work page doi:10.1137/1.9781611976700.21 2021

Show all 48 references
  1. [9]

    Yahya H Ezzeldin, Shen Yan, Chaoyang He, Emilio Ferrara, and A Salman Avestimehr. 2023. Fairfed: Enabling group fairness in federated learning. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 37. 7494– 7502

    2023

  2. [10]

    Zahra Fatemi, Chen Xing, Wenhao Liu, and Caiming Xiong. 2021. Improving gender fairness of pre-trained language models without catastrophic forgetting. arXiv preprint arXiv:2110.05367 (2021)

    2021 arXiv

  3. [11]

    van Dalen, and Matt Seigel

    Borja Rodríguez Gálvez, Filip Granqvist, Rogier C. van Dalen, and Matt Seigel

  4. [12]

    Jeremy Howard and Sebastian Ruder. 2018. Universal language model fine-tuning for text classification. arXiv preprint arXiv:1801.06146 (2018)

    2018 arXiv

  5. [13]

    Chao Jia and et al. 2021. Scaling Up Visual and Vision-Language Represen- tation Learning With Noisy Text Supervision. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2021)

    2021

  6. [14]

    Chao Jia and et al. 2022. Visual Prompt Tuning. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  7. [15]

    Junnan Li and et al. 2022. BLIP: Bootstrapping Language-Image Pre-training. arXiv preprint arXiv:2201.12086 (2022)

    2022 arXiv

  8. [16]

    Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. 2020. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems 2 (2020), 429–450

    2020

  9. [17]

    Changxin Liu, Zirui Zhou, Yang Shi, Jian Pei, Lingyang Chu, and Yong Zhang

  10. [18]

    Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017)

    2017 arXiv

  11. [19]

    CoRR abs/2109.08344 (2021)

    Achieving Model Fairness in Vertical Federated Learning. CoRR abs/2109.08344 (2021). arXiv:2109.08344 https://arxiv.org/abs/2109.08344

    2021 arXiv

  12. [20]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Agüera y Arcas

    H. Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Agüera y Arcas. 2017. Communication-Efficient Learning of Deep Networks from Decentralized Data. InProceedings of the 20th International Conference on Artificial Intelligence and Statistics. 1273–1282. ht...

    2017

  13. [21]

    Yao Lu and et al. 2023. Federated Learning with Visual Language Models: A Survey. arXiv preprint arXiv:2301.12345 (2023)

    2023 arXiv

  14. [22]

    Afroditi Papadaki, Natalia Martinez, Martin Bertran, Guillermo Sapiro, and Miguel Rodrigues. 2022. Minimax demographic group fairness in federated learn- ing. In 2022 ACM Conference on Fairness, Accountability, and Transparency. 142–159

    2022

  15. [23]

    Official Journal of the European Union. 2016. General Data Protection Reg- ulation. https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX: 32016R0679. Online; accessed 09-May-2022

    2016

  16. [24]

    Alec Radford and et al. 2021. Learning Transferable Visual Models From Natural Language Supervision. Proceedings of the International Conference on Machine Learning (ICML) (2021)

    2021

  17. [25]

    Sikha Pentyala, Nicola Neophytou, Anderson C. A. Nascimento, Martine De Cock, and Golnoosh Farnadi. 2022. PrivFairFL: Privacy-Preserving Group Fairness in Federated Learning. CoRR abs/2205.11584 (2022). https://doi.org/10.48550/arXiv. 2205.11584 arXiv:2205.11584

    2022 doi arXiv

  18. [26]

    Yuji Roh, Kangwook Lee, Steven Euijong Whang, and Changho Suh. 2020. Fair- batch: Batch selection for model fairness.arXiv preprint arXiv:2012.01696 (2020)

    2020 arXiv

  19. [27]

    Shauli Ravfogel, Yanai Elazar, Hila Gonen, Michael Twiton, and Yoav Goldberg

  20. [29]

    Xiyang Wang and et al. 2022. Investigating Gender Bias in Image Caption- ing. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  21. [30]

    Timo Schick, Sahana Udupa, and Hinrich Schütze. 2021. Self-diagnosis and self- debiasing: A proposal for reducing corpus-based bias in nlp. Transactions of the Association for Computational Linguistics 9 (2021), 1408–1424

    2021

  22. [31]

    Kellie Webster, Xuezhi Wang, Ian Tenney, Alex Beutel, Emily Pitler, Ellie Pavlick, Jilin Chen, Ed Chi, and Slav Petrov. 2020. Measuring and reducing gendered correlations in pre-trained models. arXiv preprint arXiv:2010.06032 (2020)

    2020 arXiv

  23. [32]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, et al

  24. [33]

    Kellie Webster, Marta Recasens, Vera Axelrod, and Jason Baldridge. 2018. Mind the GAP: A balanced corpus of gendered ambiguous pronouns. Transactions of the Association for Computational Linguistics 6 (2018), 605–617

    2018

  25. [34]

    Ke Yang, Charles Yu, Yi R Fung, Manling Li, and Heng Ji. 2023. Adept: A debi- asing prompt framework. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 37. 10780–10788

    2023

  26. [35]

    Qiang Yang, Yang Liu, Tianjian Chen, and Yongxin Tong. 2019. Federated machine learning: Concept and applications. ACM Transactions on Intelligent Systems and Technology (TIST) 10, 2 (2019), 1–19

    2019

  27. [36]

    Hongyu Yu and et al. 2022. CoCa: Contrastive Captioners are Image-Text Foun- dation Models. arXiv preprint arXiv:2205.01934 (2022)

    2022 arXiv

  28. [37]

    Matthew Yu Heng Wong, Nicholas YQ Tan, and Charumathi Sabanayagam. 2019. Time trends, disease patterns and gender imbalance in the top 100 most cited articles in ophthalmology.British Journal of Ophthalmology 103, 1 (2019), 18–25

    2019

  29. [38]

    Wei Zhang and et al. 2023. Learning Fair Representations with Visual Lan- guage Models. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2023)

    2023

  30. [39]

    Yao Zhang and et al. 2022. Towards Mitigating Gender Bias in Image Captioning: A Survey. arXiv preprint arXiv:2201.12345 (2022)

    2022 arXiv

  31. [40]

    Yi Zhang and Jitao Sang. 2020. Towards accuracy-fairness paradox: Adversarial example-based data augmentation for visual debiasing. In Proceedings of the 28th ACM International Conference on Multimedia. 4346–4354

    2020

  32. [41]

    Huimin Zeng, Zhenrui Yue, Yang Zhang, Lanyu Shang, and Dong Wang. 2024. Fair federated learning with biased vision-language models. In Findings of the Association for Computational Linguistics ACL 2024. 10002–10017

    2024

  33. [42]

    Wei Zhao and et al. 2018. Gender Bias in Coreference Resolution: Evaluation and Debiasing Methods. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP) (2018). Conference’17, July 2017, Washington, DC, USA Chaomeng Chen, Zitong Yu *, J...

    2018

  34. [43]

    Yao Zhao and et al. 2021. Captioning with a Focus on Gender Bias Mitiga- tion. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2021)

    2021

  35. [44]

    Jie Zhou and et al. 2022. Conditional Prompt Learning for Vision-Language Models. arXiv preprint arXiv:2204.00888 (2022)

    2022 arXiv

  36. [45]

    Haodong Zhao, Wei Du, Fangqi Li, Peixuan Li, and Gongshen Liu. 2023. Fed- prompt: Communication-efficient and privacy-preserving prompt tuning in feder- ated learning. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5

    2023

  37. [49]

    Jie Zhou and et al. 2022. Learning to Prompt for Vision-Language Mod- els. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2022)

    2022

  38. [2019]

    arXiv preprint arXiv:1910.03771 (2019)

    Huggingface’s transformers: State-of-the-art natural language processing. arXiv preprint arXiv:1910.03771 (2019)

    2019 arXiv

  39. [2020]

    arXiv preprint arXiv:2004.07667 (2020)

    Null it out: Guarding protected attributes by iterative nullspace projection. arXiv preprint arXiv:2004.07667 (2020)

    2020 arXiv

  40. [2021]

    CoRR abs/2109.08604 (2021)

    Enforcing fairness in private federated learning via the modified method of differential multipliers. CoRR abs/2109.08604 (2021). arXiv:2109.08604 https: //arxiv.org/abs/2109.08604

    2021 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.